2015 Finals: Cavs vs Warriors

[Mike G]

...
Are these adjusted for HCA ?

Yes, they are. I used HCA of 3.25 points/100 poss. They are also adjusted for SoS based on who actually played in the game (based on regular season BPM ratings)--that's why everything is positive for most offense/defense units. That doesn't affect the relative values of the players in the game vs. each other, but does affect the numbers I show, which should be valid vs. an average NBA player like BPM always is.

The expected team ratings based on lineup last night and regular season BPM: GSW +13.7 / +8.6 O / +5.1 D. CLE +5.5 / +5.5 O / +0.0 D.

I'd expect GSW+Cle 'points generated' to sum to +19.2*Pace/100, in game 3. And that is correct.
Yet in these 4 games, we have:

Code: Select all

.         Cle                  GSW                 both teams 
G   Off   Def   total     Off   Def  total      Off   Def  total
1   6.5   1.6    8.2      5.5   7.5   13.0     12.0   9.1   21.2
2   0.8  11.6   12.4     -3.6  10.7    7.2     -2.8  22.3   19.6
3   3.7   6.0    9.7      1.7   6.0    7.7      5.4  12.0   17.4
4  -3.2  -1.1   -4.3      8.8  10.7   19.5      5.6   9.6   15.2

The final column suggests that the combined productions of the Finalists are steadily eroding away.
What does it mean? Lineups should both be doing better, but they can't both do better?

[DSMok1]

Aha! It's because worse players (measured over the course of the season) are playing. Kyrie is gone. Bogut scarcely playing. So BPM sees the overall quality of the game decreasing. Last night, in particular, there was garbage time with worse players getting court time.

[Mike G]

Warriors 3, Cavs 2
avg score 99.8 - 92.8
pyth wins 3.62 - 1.38, eWins 3.06 - 1.94

Code: Select all

eWins   per36 rates   tm   mpg   Eff%   Sco   Reb   Ast    PF   Stl   TO   Blk   e500
1.35   LeBron James   Cle   46  .475   28.0  10.0   6.4   2.5   1.0   2.8   .4   2.97
.88   Stephen Curry   GSW   42  .585   27.3   4.3   5.7   1.7   1.4   4.4   .2   2.07
.52   Andre Iguodala  GSW   37  .608   16.7   5.8   4.2   2.7   1.0   1.2   .4   1.39
.51   Draymond Green  GSW   36  .477   13.0   7.9   4.7   4.7   2.0   2.1  1.2   1.42
.38   Timofey Mozgov  Cle   27  .588   20.7   9.0    .9   2.7    .8   3.8  1.2   1.39

.36   Klay Thompson   GSW   40  .517   15.6   3.8   1.7   2.5    .5    .6   .9    .89
.25  Tristan Thompson Cle   42  .500    7.1  11.4    .3   2.4    .7    .8   .8    .60
.19   Kyrie Irving    Cle    9  .511   19.2   6.0   4.8   4.2   3.4   1.0  1.6   2.26
.19   David Lee       GSW    7  .564   24.2   9.2   6.3   7.3    .9    .9   .0   2.60
.15   Harrison Barnes GSW   32  .463    9.0   7.5   1.1   3.2    .9    .9   .7    .47

eWins   per36 rates   tm   mpg   Eff%   Sco   Reb   Ast    PF   Stl   TO   Blk   e500
.12  Shaun Livingston GSW   18  .455    8.0   7.0   5.9   3.1    .8   2.4   .4    .68
.12   Andrew Bogut    GSW   15  .455    4.9  11.8   3.4   5.4    .5   2.0  2.5    .78
.10   Leandro Barbosa GSW   10  .565   20.2   4.7   1.6   4.0    .7   2.8   .7   1.03
.07   Festus Ezeli    GSW    8  .500   11.4  11.5   2.1   6.2    .0   2.7  1.8    .92
.04 Marreese Speights GSW    3  .321   14.8  10.4   2.5   6.2   2.1   2.1   .0   1.32

.01   J.R. Smith      Cle   34  .397    7.8   4.5   1.0   3.4    .4    .8   .8    .02
-.05   Mike Miller    Cle    6  .750    4.2   1.2    .0   3.6   1.2   1.4   .0   -.90
-.06   Iman Shumpert  Cle   36  .388    4.8   4.1    .8   2.7   2.2   1.2   .6   -.16
-.06 Matt Dellavedova Cle   33  .404    8.3   2.7   2.9   2.8   1.5   3.4   .0   -.19
-.08   James Jones    Cle   17  .517    5.6   2.9    .7   6.4   1.6   1.4   .4   -.45

shooting: GS .527 - .469 Cle
rebounds: Cle 44.0 - 43.4 GS
turnovers: Cle 13.6 - 13.8 GS

[DSMok1]

Single Game Box Plus/Minus:



LeBron's trying, but... he isn't getting much help.

And here are the totals through the first 5 games in the series:

[Mike G]

Thanks for the series totals.
In the OP, a table starts like this:

Code: Select all

Cavs     Min     PER    e480   eWins    WS    WS/48    BPM  VORP
LeBron   569    24.8    2.91   3.42    2.15    .181    9.7   1.7
...

Your table above says LeBron's BPM for the Finals is 10.7 -- up from 9.7 thru the ECF.
Yet on b-r.com, the Cavs page lists his playoff BPM to date as 11.0, for 798 min.
http://www.basketball-reference.com/tea ... eries_team
How can a 10.7 BPM (in 228 min.) raise a 9.7 to 11.0 ?

By my calculations, it should be : (798*11.0 - 569*9.7)/229 = 14.2 BPM for Finals (G 1-5)
Pace would be a small factor, but: Players on both teams have discrepancies large and small, positive and negative.
Mozgov was 0.7 coming in. After 5 games at 2.3, he's now at 0.4 ?

[DSMok1]

Single game BPM is not the full BPM computation, and doesn't include the nonlinear terms (so it will stay more stable in small sample sizes like here). That's the difference.

[Mike G]

Here's everyone with 20+ minutes in the series. eWins and eW/484 are calculated exclusively for this series; PER, WS, BPM, and VORP are implied/inferred from: (total playoff minutes and rates) - (pre-finals min. and rates). <-- see the OP
Some variation will be expected from pace changes -- GSW have slowed down more in the Finals.

Code: Select all

Cavs     Min    PER   e484  eWins   WS  WS/48   BPM  VORP
LeBron   228   27.7   2.97  1.35   .86   .182  14.3   .93
Thompson 209   13.9    .60   .25   .53   .121    .0   .10
Shumpert 179    6.3   -.16  -.06   .02   .005   -.1   .08
Smith    168    5.7    .02   .01  -.04  -.012  -3.5  -.06
Mozgov   136   19.1   1.39   .38   .36   .125   -.4   .05

Dellav'a 166    4.9   -.19  -.06  -.20  -.057  -4.2  -.09
Jones     85    4.1   -.45  -.08   .05   .030  -2.1   .00
Irving    43   26.8   2.26   .19   .13   .141   8.5   .11
Miller    28    2.4   -.90  -.05   .01   .022  -3.8  -.01

Cavs    1250   13.1    .77  1.93  1.71   .066  1.45  1.08
                           
                           
GSW      Min    PER   e484  eWins   WS  WS/48   BPM  VORP
Curry    212   20.1   2.07   .88   .68   .153   6.4   .44
Thompson 202   14.1    .89   .36   .56   .134   2.0   .20
Iguodala 186   16.9   1.39   .52   .75   .194   5.8   .36
Green    180   15.5   1.42   .51   .63   .167   5.5   .34
Barnes   162   11.6    .47   .15   .42   .123   2.6   .19

Living'n  88   10.6    .68   .12   .18   .096  -1.0   .02
Bogut     74   10.5    .78   .12   .21   .138   2.6   .09
Barbosa   50   14.1   1.03   .10   .11   .104  -2.0   .00
Ezeli     38   12.4    .92   .07   .10   .132  -1.5   .00
Lee       37   24.4   2.60   .19   .24   .312   7.4   .09

GSW     1248   15.1   1.23  3.06  3.84   .148  3.42  1.69

Win Shares sum to 5.55 after 5 games. PER look like they should all be 0.9 higher (to sum to 15)

EDIT: 'Single game BPM' has been dropped from this table. Those shown are derived from formula shown a couple of posts up. Based on current BPM showing at b-r.com and BPM/minutes shown before the Finals.

[Mike G]

Single game BPM is not the full BPM computation, and doesn't include the nonlinear terms (so it will stay more stable in small sample sizes like here). That's the difference.

Ah. So the 5-game totals are just the 5 single games, totaled?

[DSMok1]

Yes.

[Mike G]

Is there any reason not to derive single series (or game) BPM the way I've done it (and PER, and WS/48) from the larger sample, just by subtracting the minutes * rates before the ongoing series?
Do they not add up that way?

If you've got a series of 'single game BPM' that do not add up to the postseason BPM we see at b-r.com, are these BPM even legit? It seems the more stable one is the one you don't have to ignore after the totals are in. Is Steph Curry running a SGBPM of 3 to 5 in other series as well? Can these give him a total playoff BPM of 8.8 ?

One implication in the dueling BPM systems: As I've derived it, LeBron is the only member of the Cavs with positive BPM in the Finals (after game 1 Irving). He has produced (93/108) 86% of the team's Value (ORP) -- and 96% of the non-Irving VORP.

He also has more than twice the VORP of any member of the Warriors. Curry has .44 vs LeBron's .93
Other stats concur, it's a landslide at this point.
Can a loser win Finals MVP ? Can he not?

[DSMok1]

Is there any reason not to derive single series (or game) BPM the way I've done it (and PER, and WS/48) from the larger sample, just by subtracting the minutes * rates before the ongoing series?
Do they not add up that way?

If you've got a series of 'single game BPM' that do not add up to the postseason BPM we see at b-r.com, are these BPM even legit? It seems the more stable one is the one you don't have to ignore after the totals are in. Is Steph Curry running a SGBPM of 3 to 5 in other series as well? Can these give him a total playoff BPM of 8.8 ?

One implication in the dueling BPM systems: As I've derived it, LeBron is the only member of the Cavs with positive BPM in the Finals (after game 1 Irving). He has produced (93/108) 86% of the team's Value (ORP) -- and 96% of the non-Irving VORP.

He also has more than twice the VORP of any member of the Warriors. Curry has .44 vs LeBron's .93
Other stats concur, it's a landslide at this point.
Can a loser win Finals MVP ? Can he not?

In very small sample sizes, the interaction terms of the full BPM can produce unrealistic outlier solutions. The way you are doing it is as good as any, but I don't know how stable the regression is for numbers like LeBron's--no one in my long term RAPM regression had box score numbers approaching what he is doing. Therefore, I'd trust a more simplified, linear regression more for such a case. Both have value as two different looks at the same data, neither is "right."

[Mike G]

In very small sample sizes, the interaction terms of the full BPM can produce unrealistic outlier solutions. ..

We all get that. :]
Let me ask in a slightly different way: Why not use the considerably larger sample size based on 19-20 games; subtract out the considerable sample size of the first 14-15; and get a BPM that's (a) more stable AND (b) consistent with the BPM shown at b-r.com? Seems like a win-win.

For what it's worth, both Win Shares and eWins correlate better with extracted BPM than with the streamlined version, in this series. Minutes are better correlated -- with or without Irving in the sample -- even with the Cavs giving big minutes to poor (by every measure) players.

[Mike G]

Other than LeBron and Irving, the Cavs in the Finals have total PER = 9.1, eW/484 = .20, WS/48 = .035, BPM = -1.85
Basically Mozgov and replacement types.

[Mike G]

Warriors 4, Cavs 2
avg score 100.7 - 93.5
Pyth wins 4.37 - 1.63, eWins 3.68 - 2.32

Code: Select all

eWins   per36 rates   tm   mpg   Eff%   Sco   Reb   Ast    PF   Stl   TO   Blk   e497
1.57   LeBron James   Cle   46  .468   27.1  10.7   6.6   2.5   1.1   3.3   .4   2.84
1.07   Stephen Curry  GSW   43  .578   26.8   4.3   6.2   1.9   1.6   4.1   .1   2.09
.70   Draymond Green  GSW   37  .488   13.6   8.0   5.7   4.6   2.1   2.2  1.2   1.56
.69   Andre Iguodala  GSW   37  .576   18.1   5.5   4.4   2.9   1.1   1.0   .3   1.53
.54   Timofey Mozgov  Cle   28  .609   21.4   9.9   1.1   2.8    .6   3.9  1.8   1.58

.37   Klay Thompson   GSW   38  .505   15.0   4.1   1.9   3.2    .5    .8  1.0    .80
.34  Tristan Thompson Cle   41  .517    8.3  11.7    .3   2.4    .6    .9   .8    .69
.19   Kyrie Irving    Cle    7  .511   19.3   6.0   4.8   4.2   3.4   1.0  1.6   2.21
.17   David Lee       GSW    7  .537   21.6   8.3   5.8   6.8    .9    .9   .0   2.14
.15   Harrison Barnes GSW   33  .477    9.1   6.3   1.1   3.2    .9    .8   .6    .37

eWins   per36 rates   tm   mpg   Eff%   Sco   Reb   Ast    PF   Stl   TO   Blk   e497
.14   Leandro Barbosa GSW   11  .554   18.2   5.0   1.8   3.6   1.6   2.1   .5   1.08
.13   Festus Ezeli    GSW    8  .537   16.9  11.2   1.6   6.1    .0   2.8  2.1   1.35
.13  Shaun Livingston GSW   20  .517    9.2   5.8   4.3   3.1    .6   1.7   .6    .54
.11   Andrew Bogut    GSW   12  .455    4.9  11.4   3.4   5.4    .5   2.0  2.5    .73
.08   J.R. Smith      Cle   34  .421    9.4   4.6   1.0   3.3    .5    .8   .8    .19

.04 Marreese Speights GSW    3  .321   14.5   9.9   2.4   6.1   2.0   2.1   .0   1.20
-.05   Mike Miller    Cle    5  .750    4.1   1.2    .0   3.5   1.2   1.4   .0   -.88
-.10   Iman Shumpert  Cle   36  .390    5.2   4.0    .6   3.2   1.9   1.4   .6   -.24
-.12 Matt Dellavedova Cle   32  .385    7.2   3.1   2.9   3.6   1.3   3.4   .0   -.31
-.13   James Jones    Cle   19  .476    5.3   2.5    .6   5.7   1.2   1.4   .3   -.59

shooting: GS .528 - .470 Cle
rebounds/48: Cle 45.9 - 42.7 GS
turnovers: GS 13.0 - 14.5 Cle

[Mike G]

Again extracting PER, WS, and BPM for these last 6 games -- Finals rates and totals:

Code: Select all

GSW       Min    PER   e484  eWins    WS   WS/48   BPM  vorp
Curry     255   20.2   2.09  1.07    .83   .156    6.5   .54
Thompson  227   12.2    .80   .37    .50   .106    1.3   .19
Iguodala  223   18.0   1.53   .69    .92   .198    5.8   .43
Green     222   16.3   1.56   .70    .81   .176    6.3   .46
Barnes    197   10.7    .37   .15    .47   .115    2.1   .20

Living'n  121   10.9    .54   .13    .29   .116    -.3   .05
Bogut      74   10.5    .73   .11    .21   .138    2.6   .09
Barbosa    64   15.6   1.08   .14    .17   .126     .0   .03
Ezeli      49   17.5   1.35   .13    .19   .186   -1.1   .01
Lee        39   21.9   2.14   .17    .22   .271    5.0   .07

GSW      1490   15.2   1.24  3.69   4.59   .148   3.44  2.03
 
                       
Cavs      Min    PER   e484  eWins    WS   WS/48   BPM  vorp
LeBron    275   26.3   2.84  1.57    .90   .156   13.7  1.08
Thompson  247   14.4    .69   .34    .63   .122    -.3   .10
Shumpert  215    5.5   -.24  -.10   -.01  -.003    -.9   .06
Smith     202    7.3    .19   .08    .06   .015   -2.8  -.04
Dellav'a  191    3.4   -.31  -.12   -.26  -.067   -4.8  -.13

Mozgov    169   21.5   1.58   .54    .59   .167    2.3   .18
Jones     113    2.8   -.59  -.13    .03   .012   -3.5  -.04
Irving     43   26.8   2.21   .19    .12   .131    8.5   .11
Miller     28    2.4   -.88  -.05    .01   .020   -4.2  -.02

Cavs     1490   13.0    .78  2.31   2.04   .066   1.39  1.26

Win Shares likes Iguodala for Finals mvp!